[BUG] Overflows in irls_solver and shape mismatch in _refit_without_outliers when using continuous variable

When using a continuous variable I get
ValueError: shape mismatch: value array of shape (228,77) could not be broadcast to indexing result of shape (228,7)
while running self.varm["LFC"][to_replace] = sub_dds.varm["LFC"] in _refit_without_outliers.

I also get overflow warnings from irls_solver.

• Setting 1
  design_factors=['Batch','Condition']
  continuous_factors=None

Works fine

• Setting 2
  design_factors=['Batch', 'Age', 'Condition']
  continuous_factors=None

Works fine (Refitting 29547 outliers, compared to 1567 in the previous setting, does this make sense?)

• Setting 3
  design_factors=['Batch', 'Age', 'Condition']
  continuous_factors=['Age']

Overflows in irls_solver (similar to Issue #169). To be precise, I do not get these in previous settings.

ValueError: shape mismatch: value array of shape (228,77) could not be broadcast to indexing result of shape (228,7)
(while running self.varm["LFC"][to_replace] = sub_dds.varm["LFC"] in _refit_without_outliers)

Reproduce the issue using the three design_factors and continuous_factors settings in the report, then inspect _refit_without_outliers where self.varm["LFC"] is assigned and irls_solver where the overflow warnings occur. Done means the continuous-variable case no longer raises the reported shape-mismatch error and the associated overflows are addressed.